1314741 - [User Story] As a user, I understand and expect to find the same AS items across all devices I synced Firefox to.

Status: RESOLVED FIXED

(Firefox for Android Graveyard :: General, defect)

Comment 1

[:Grisha Kruglov]

Similarity of the "stream" portion of the Activity Stream is a function of how Sync functions. As long as devices are syncing bookmark and history collections with relative frequency, the underlying data will be there, and experiences will be consistent.

Then there's a question of two additional sections: Top Sites and Highlights.

Highlights algorithm work and telemetry are addressed as part of this user story: Bug 1314724.

Top Sites deserves further thought, I think. We've done the work necessary to track whether or not visits are locally generated (from Fennec), or came in from other Sync clients. This work was done so that we can either exclude desktop history from Top Sites, or heavily prioritize mobile history over desktop history. Relevant bugs and discussions: meta Bug 1265516 tracks various parts of this work, and Bug 1278301 pulls the trigger on desktop visits.

However, we've never shipped the final UI change. Bug 1278301 was backed out, and at the time it was decided that we don't have enough quantifiable data about Top Sites usage to make a decision that excluding or de-prioritizing remote history is the way to go. It was decided to wait until there are more resources available to push the necessary telemetry work through. That's the background, and I believe the same pros and cons from those discussions apply to AS Top Sites.

While consistency across platforms is important, I think doing so at the expense of usability of data presented is not a good idea. If AS Top Sites are thought of "my top sites in relation to my browsing medium", then we can achieve consistency in experience by ensuring that we display in Top Sites is in fact what the user considers their Top Sites to be for mobile, desktop, etc.

This could be the same as "make our top sites algorithms the same across platforms", but probably not, since I don't believe desktop currently tracks source of visits (neither does mobile, "local vs remote visit" is a bit of a hack to get around sync limitations).

Barbara, what do you think about this?

Comment 2

[Richard Newman [:rnewman]]

I'm not sure this is (a) possible, (b) desirable.

The possible part:

- The two devices very likely have different data. Your desktop probably has much more history than your phone. Even if you have two desktops, they very likely don't have the same history.

- Your desktops will have container history (but not synced!). Mobile devices don't know about containers: indeed, they probably won't even see non-default container history at all.

- Even if you have all of the same history items, your desktop has more visit metadata that simply isn't synced.

- We've been discussing adding _even more data_ like dwell time, which also isn't present in Sync.

- Desktop and mobile frecency algorithms are not the same.


The desirable part:

We spent quite a lot of effort trying to make sure that your local top sites on mobile devices are _not_ the same as your top sites on other devices, because otherwise your desktop browsing habits overwhelm your mobile device (or vice versa, for some users, and perhaps that's OK?).

That was predicated on the assumption that one's mobile and desktop patterns of use are not the same: perhaps I read photography blogs on desktop, and on my phone I look at recipes and cat pictures.

I expect this care to extend to all aspects of the AS experience: the bookmarks we surface, the history we emphasize, the suggestions we make, and the content you might want to revisit should all be contextualized appropriately and deliberately. Yes, perhaps encourage me to re-read something I saw on my desktop… but don't blindly make the same suggestions in both places.

Comment 3

[Barbara Bermes [:barbara] - NI please!]

All good points and I'd like to get Alex D. opinion on this as well it relates to not only AS but sync in general, and what he has envisioned for his sync product.

As for "Desktop and mobile frecency algorithms are not the same", is this intentional?

I wonder if labeling items differently would help, e.g. little mobile icon indicates you've viewed this only on Mobile etc.

Maybe we need/should do some UR around this, as we all just guess what the user's expectations and use cases are. 

In addition, Grisha, how about we start with making it the same in Nightly, and keep a close eye on any telemetry like https://bugzilla.mozilla.org/show_bug.cgi?id=1318550 that we can gather for Top Sites and Highlights.

Comment 4

[:Grisha Kruglov]

(In reply to Barbara Bermes [:barbara] - NI please! from comment #3)
> All good points and I'd like to get Alex D. opinion on this as well it
> relates to not only AS but sync in general, and what he has envisioned for
> his sync product.
> 
> As for "Desktop and mobile frecency algorithms are not the same", is this
> intentional?
> 
> I wonder if labeling items differently would help, e.g. little mobile icon
> indicates you've viewed this only on Mobile etc.
> 
> Maybe we need/should do some UR around this, as we all just guess what the
> user's expectations and use cases are. 
> 
> In addition, Grisha, how about we start with making it the same in Nightly,
> and keep a close eye on any telemetry like
> https://bugzilla.mozilla.org/show_bug.cgi?id=1318550 that we can gather for
> Top Sites and Highlights.

I'm not sure it's worth the effort trying to make our Top Sites sections display exact same sets of items across devices. Our algorithms are different (mobile ones are an "approximation" of real frecency, for performance/sqlite limitations reasons), and even if they were the same, underlying data might/will be different enough so that generated top site items will vary - see Comment 2.

My vote would be to:
- implement the mentioned telemetry so that we know how people are interacting with Top Sites
- consider implementing platform-specific top sites/highlights for AS MVP as an experiment
- and couple that with UR to validate our assumptions about users' browsing habits as they move from platform to platform.

Comment 5

[Richard Newman [:rnewman]]

(In reply to Barbara Bermes [:barbara] - NI please! from comment #3)

> As for "Desktop and mobile frecency algorithms are not the same", is this
> intentional?

Yes: the mobile ones are simpler, and the strategy for frecency updates differs because there's no such thing as idle-time on mobile, so we have no opportunity for decay.

I expect both mobile and desktop frecency algorithms to change over the next year as they adapt to changing storage infrastructure, additional data (e.g., compartments and dwell time), and computational resources.

(But note that even if the algorithms were the same, as they are between two desktop browsers, they'd still produce different results for the different database contents on each machine. In order to produce the same results, we'd need to sync much more data than we do now.)

 
> I wonder if labeling items differently would help, e.g. little mobile icon
> indicates you've viewed this only on Mobile etc.

It might, but note that we don't sync that data: Bug 1302797. The best we can do on mobile is "did you view it on this device or not?", and false negatives are possible. Desktop Firefox can't do this at all yet, AFAIK.

Comment 6

[Alex Davis [:adavis]]

Hi All,

According to nchapman, they are not suppose to be the same. In fact, the Sync team is working on 2 things in Q4 for AS to make sure that mobile and desktop always show the most relevant content and so that AS can know where content came from:
- Add device origin to sync'd bookmarks Bug 1302797
- Add creation date. Currently, the bookmark creation date on a 2nd device is when it was synced and not when it was created on original device. (can't find bug at this time)

Hope this helps

Comment 7

[Richard Newman [:rnewman]]

(In reply to Alex Davis [:adavis] from comment #6)

> - Add creation date. Currently, the bookmark creation date on a 2nd device
> is when it was synced and not when it was created on original device. (can't
> find bug at this time)

Bug 676563.

Comment 8

[Barbara Bermes [:barbara] - NI please!]

> My vote would be to:
> - implement the mentioned telemetry so that we know how people are
> interacting with Top Sites
> - consider implementing platform-specific top sites/highlights for AS MVP as
> an experiment
> - and couple that with UR to validate our assumptions about users' browsing
> habits as they move from platform to platform.

Ok, let's go with that.

And as per comment 6, Alex, I assume the front-end team will need to change their algorithm accordingly based on the additional meta data.

Also, I want to make sure that iOS and Android consume/deal with it the same way.

I want to make sure no unnecessary work is being done right now (if things change again)

Comment 9

[:Grisha Kruglov]

I've left a clarifying comment on the iOS counterpart of this work (Bug 1314384).
Filed Bug 1321368 to prioritize local history (as we did before for regular Top Sites).
Re-purposed Bug 1318550 to track work on sync-aware telemetry of interactions with Top Sites and Highlights.